Asphalt mixing process dynamic optimization method based on multi-data fusion and digital twinning

Abstract

The invention discloses an asphalt mixing process dynamic optimization method based on multi-data fusion and digital twinning, and relates to the field of road engineering, and the method comprises the steps: collecting multi-source sensor data in the production process of an asphalt mixing station, carrying out the fusion processing of the multi-source sensor data, and obtaining a state parameter feature set, the multi-source sensor data comprises temperature, humidity, flow, vibration signals, dust concentration and asphalt viscosity; inputting the state parameter feature set into the intelligent optimization model, and outputting an optimized process parameter set value through iterative search and prediction feedback; and the optimized process parameter set value is issued to an asphalt mixing station control system for execution, and a digital twin model is driven based on real-time production data to carry out visual monitoring and closed-loop control. By collecting multi-source data and combining an intelligent optimization algorithm, automatic adjustment and real-time quality prediction of process parameters are achieved, the intelligent level of the production process is improved, stable product quality is ensured, and dynamic working condition changes are coped with.

Description

Technical Field

[0001] This invention relates to the field of road engineering, specifically to a dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twins. Background Technology

[0002] Asphalt mixture mixing is a critical step in road construction, and its quality directly affects pavement performance. This process involves multiple stages, including drying, metering, and mixing, and is a multivariable, strongly coupled, and nonlinear process dynamically influenced by raw material characteristics, equipment status, and environmental factors. Traditional production relies mainly on fixed parameter settings and manual experience, making it difficult to adapt in real-time to fluctuations in key parameters such as aggregate moisture content and asphalt viscosity, resulting in unstable product quality and high energy consumption.

[0003] In existing technologies, while local optimization methods based on a single model can improve some control accuracy, they generally suffer from weak generalization ability, susceptibility to local optima, and difficulty in capturing long-term temporal dependencies when dealing with full-process, multi-objective optimization. Furthermore, the lack of deep integration and intuitive display of production data and physical equipment status limits operators' perception of the overall production situation and decision-making efficiency, thus hindering further improvements in production quality. Summary of the Invention

[0004] Based on the shortcomings of the prior art described above, the purpose of this invention is to provide a dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twins, so as to solve the above-mentioned technical problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twin, comprising:

[0006] Multi-source sensor data are collected during the production process of asphalt mixing plant, and the multi-source sensor data is fused to obtain a set of state parameter features. The multi-source sensor data includes temperature, humidity, flow rate, vibration signal, dust concentration and asphalt viscosity.

[0007] The state parameter feature set is input into the intelligent optimization model, and the optimized process parameter set value is output through iterative search and prediction feedback.

[0008] The optimized process parameter settings are sent to the asphalt mixing plant control system for execution, and the digital twin model is driven by real-time production data for visual monitoring and closed-loop control.

[0009] The present invention is further configured such that the fusion processing of multi-source sensor data includes:

[0010] An extended Kalman filter is used to filter the acquired continuous variable sensor data to eliminate high-frequency noise;

[0011] Principal component analysis was performed on the raw sensor data to extract the top K principal components whose cumulative variance contribution rate was greater than a preset contribution rate threshold.

[0012] When the difference in readings between data from the same type of sensor exceeds a preset difference threshold, a data fusion algorithm is used to process the data and obtain consistent data.

[0013] Based on historical production data, a correlation analysis of state parameters is conducted.

[0014] The present invention is further configured such that the data fusion algorithm is used for processing, specifically including:

[0015] A basic probability assignment function is constructed for each data source with reading differences. The assignment of the basic probability assignment function is based on the historical measurement accuracy of the data source and the degree of deviation of its current reading from the statistical characteristics of the reading group of other data sources in the same period.

[0016] Based on the DS evidence theory, the basic probability assignment function is synthesized and calculated, and the consistency data is determined according to the maximum confidence proposition of the synthesis result.

[0017] The present invention is further configured such that the state parameter correlation analysis includes:

[0018] Based on historical production data, the Pearson correlation coefficients between each pair of state parameters collected by the sensors are calculated to form a correlation matrix;

[0019] When the absolute value of the Pearson correlation coefficient between any two state parameters in the correlation matrix is ​​greater than the preset correlation threshold, it is determined that there is a strong correlation between the two state parameters.

[0020] The present invention is further configured such that the intelligent optimization model is a collaborative model of particle swarm optimization algorithm and long short-term memory network model;

[0021] The initialization of the particle swarm optimization algorithm includes: setting up a population containing a preset number of particles, where the position vector of each particle represents a combination of process parameters, and setting a maximum number of iterations for the algorithm. The combination of process parameters includes the aggregate mix ratio, mixing time, dry drum burner temperature, and asphalt injection amount for each cold aggregate bin.

[0022] The initialization of the Long Short-Term Memory Network model includes: configuring its input layer to receive time-series data organized by sliding time windows of a specific time length, and configuring its output layer to predict key quality indicators of future batches of asphalt mixtures. The time-series data includes historical process parameters and real-time operating condition data, and the key quality indicators include asphalt content, key sieve pass rate, porosity, and stability.

[0023] The collaboration is achieved iteratively, with each iteration including:

[0024] The particle swarm optimization algorithm generates a set of candidate process parameters based on the current population state;

[0025] Each candidate process parameter set and real-time operating data are input into a long short-term memory network model to predict the key quality indicators of the asphalt mixture produced when the candidate process parameter set is applied.

[0026] The particle swarm optimization algorithm updates particle velocity, position, and the swarm's historical best position based on the deviation between the predicted key quality indicators and the target value.

[0027] The iteration process continues until the preset termination condition is met, and the combination of process parameters corresponding to the group's historical best position is output as the optimized process parameter set value.

[0028] The present invention is further configured such that the digital twin model is a three-dimensional model of an asphalt mixing plant constructed based on building information modeling technology, and the digital twin model accurately maps the overall layout of the plant area and the structure of key equipment;

[0029] The digital twin model provides a detailed model of the mixing plant, drying drum, and dust removal system. It can display the flow state of materials inside the mixing tank and weighing bin, and dynamically simulate the rotational motion of the drying drum and the trajectory of the aggregate.

[0030] Each component in the digital twin model is associated with equipment attribute information and is deeply integrated and dynamically mapped with real-time production data to achieve three-dimensional visualization monitoring of production status.

[0031] The present invention is further configured such that the visualization monitoring includes process parameter transparency, production status visualization, and carbon emission visualization:

[0032] The process parameters are made transparent by displaying aggregate ratio, asphalt dosage, mixing time, inlet and outlet temperatures and rotation speed in real time on the corresponding equipment in the digital twin model;

[0033] The production status visualization is displayed on the plant layout as a heat map showing the distribution of global energy consumption or production efficiency.

[0034] The carbon emission visualization is based on real-time collected production data, dynamically calculates instantaneous carbon emissions, and simulates the carbon emission trajectory in a digital twin model in the form of particle flow, visually displaying the emission source to the environment.

[0035] The present invention is further configured such that the closed-loop control is implemented through a rule-based three-level alarm mechanism:

[0036] Level 1 warning: Triggered when process parameters deviate from the preset optimal value but are still within the preset safety range, the relevant equipment or area is highlighted in blue in the digital twin model;

[0037] Level 2 Alarm: Triggered when process parameters are greater than or equal to a preset safety threshold, automatically generating an early warning prompt and initiating the maintenance work order generation process;

[0038] Level 3 alarm: Triggered when process parameters are greater than or equal to preset danger thresholds, the relevant equipment or area will be highlighted and flashed red in the digital twin model, and automatic interlock control will be implemented to reduce equipment load or start emergency shutdown procedures.

[0039] The present invention is further configured such that the closed-loop control is specifically implemented in an adaptive adjustment process for abnormal asphalt viscosity conditions, including:

[0040] When the online viscometer detects that the asphalt viscosity value exceeds the preset optimal range for a consecutive preset number of sampling periods, the particle swarm optimization algorithm is automatically triggered to use the current asphalt viscosity as the optimization boundary condition, maintain the target porosity as the optimization objective, and recalculate the optimized process parameter settings.

[0041] Based on the recalculated optimized process parameter settings, adjustment instructions are sent to the mixing plant control system.

[0042] The present invention is further configured such that the adjustment command includes a compensation value for the temperature of the dryer drum burner and an extension value for the wet mixing time of the mixing tank.

[0043] This invention provides a dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twin. It collects multi-source sensor data during the asphalt mixing plant production process and fuses this data to obtain a set of state parameter features. The multi-source sensor data includes temperature, humidity, flow rate, vibration signal, dust concentration, and asphalt viscosity. The state parameter feature set is input into an intelligent optimization model, which outputs optimized process parameter setpoints through iterative search and predictive feedback. These optimized process parameter setpoints are then sent to the asphalt mixing plant control system for execution. Furthermore, the method uses real-time production data to drive a digital twin model for visualized monitoring and closed-loop control. The resulting benefits include:

[0044] 1. By collecting and fusing multi-source sensor data in real time and combining it with intelligent optimization algorithms, process parameters can be adjusted in real time to automatically adapt to various dynamic changes in the production process, thereby improving the automation and intelligence of the production process, reducing human intervention, and ensuring the consistency of product quality and the stability of the process.

[0045] 2. By adopting a collaborative optimization approach combining particle swarm optimization algorithm and long short-term memory network model, the quality of asphalt mixtures can be fed back and predicted in real time during the production process. The optimized process parameters can effectively cope with changes in working conditions, improve the stability and consistency of the process, and reduce quality fluctuations caused by operational errors or environmental changes.

[0046] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0048] Figure 1 The flowchart illustrates an exemplary embodiment of the present invention, showing a dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twin. Detailed Implementation

[0049] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0050] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0051] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0052] A dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twin, such as Figure 1 As shown, it includes:

[0053] Multi-source sensor data are collected during the production process of asphalt mixing plant, and the multi-source sensor data is fused to obtain a set of state parameter features. The multi-source sensor data includes temperature, humidity, flow rate, vibration signal, dust concentration and asphalt viscosity.

[0054] The state parameter feature set is input into the intelligent optimization model, and the optimized process parameter set value is output through iterative search and prediction feedback.

[0055] The optimized process parameter settings are sent to the asphalt mixing plant control system for execution, and the digital twin model is driven by real-time production data for visual monitoring and closed-loop control.

[0056] The present invention is further configured such that the fusion processing of multi-source sensor data includes:

[0057] An extended Kalman filter is used to filter the collected continuous variable sensor data to eliminate high-frequency noise. Specifically, in the asphalt mixing production process, continuous variable data such as temperature and flow rate collected by sensors are easily affected by the complex on-site environment and contain high-frequency random noise. This noise seriously affects the data quality and the reliability of subsequent analysis. This invention uses an extended Kalman filter to filter such data. At each sampling time, the algorithm first performs a state prediction step, that is, based on the optimal state estimate value of the previous time and the system input, it calculates the prior prediction value of the current state and its uncertainty measure through a nonlinear state transition equation. Then, a measurement update step is performed, whereby the algorithm obtains the actual observation value of the sensor and compares the observation value with the prior prediction value. The algorithm adaptively weights and fuses the two prediction models based on their respective uncertainty covariance matrices. This weighting process is specifically reflected in the calculation of the Kalman gain: when the prediction model has high reliability but the observation noise is significant, the Kalman gain decreases, and the algorithm favors the predicted value, thus effectively suppressing abrupt noise in the observation value; conversely, when the observation accuracy is high but the prediction uncertainty increases, the Kalman gain increases, and the algorithm focuses more on the observation value to track the actual state changes. Finally, the algorithm synthesizes the state prediction value with the Kalman gain-weighted information and outputs the optimized posterior state estimate at the current time, which is the result after filtering and denoising. Through the above continuous prediction and correction iterations, the extended Kalman filter can extract smooth and reliable process data from the noise-contaminated original signal.

[0058] Principal component analysis (PCA) is performed on the raw sensor data to extract the top K principal components whose cumulative variance contribution rate exceeds a preset contribution rate threshold. Specifically, to extract features representing the core state of the process from raw sensor data with overlapping information, PCA is used to perform dimensionality reduction and feature enhancement. The process is as follows: raw observation data with different dimensions collected from multiple sensors within a specific time window are constructed into a multidimensional observation vector as input. A new set of orthogonal bases is found through orthogonal linear transformation, with the directions being the projection directions with the largest data variance. The algorithm first calculates the first principal component that can explain the overall variance of the raw data to the greatest extent, then calculates the second principal component that is orthogonal to the previous components and can explain the remaining variance to the greatest extent, and so on. The extraction of principal components continues until the cumulative variance contribution rate of the selected top K principal components exceeds the preset contribution rate threshold, indicating that these principal components have covered most of the variation information of the raw data. Then, the score vector of the selected principal components is used to replace the original high-dimensional sensor readings as the state parameter feature set representing the core state of the production process, thereby reducing the data dimensionality and eliminating redundancy while retaining key information to the maximum extent.

[0059] When the difference in readings between data from the same type of sensors exceeds a preset difference threshold, a data fusion algorithm is used to process the data and obtain consistent data. The invention further specifies that the data fusion algorithm processing includes: constructing a basic probability allocation function for each data source with reading differences. The assignment of the basic probability allocation function is based on the historical measurement accuracy of the data source and the degree of deviation of its current reading from the statistical characteristics of other data source readings during the same period; performing a synthesis calculation on the basic probability allocation function based on the DS evidence theory, and determining the consistent data based on the maximum confidence proposition of the synthesis result. Specifically, to solve the decision-making problem when monitoring data from the same type of sensors are inconsistent, a conflict data fusion method based on DS evidence theory is adopted. The specific implementation process is as follows: when the difference between multiple sensor readings for the same physical quantity is detected to exceed a preset difference threshold, the fusion process is triggered. First, for each data source with reading differences... The data sources are used to construct basic probability allocation functions. The values ​​of these functions are mainly based on the historical measurement accuracy of the data source and the degree of deviation of its current reading from the statistical characteristic values ​​of other conflicting data source readings during the same period. The historical measurement accuracy is quantified by converting the average relative error or root mean square error between the sensor's periodic readings and the laboratory reference value into a normalized confidence score through a mapping function. The degree of deviation of the current reading is quantified by calculating its relative deviation from the statistical characteristic values ​​of the conflicting data group during the same period. The statistical characteristic values ​​can be the median or the truncated mean. The basic probability allocation functions of each data source are used as independent evidence. The evidence is combined according to the synthesis rules of DS evidence theory. These rules calculate the joint support of all evidence for the same proposition and quantify the conflict within the evidence set. The synthesis calculation produces a new basic probability allocation that reflects the combined effect of all evidence. Based on this, the data value corresponding to the proposition with the highest confidence is selected as the consistent fusion result output.

[0060] Based on historical production data, a correlation analysis of state parameters is performed. The invention further specifies that the correlation analysis includes: calculating the Pearson correlation coefficient between each pair of state parameters collected by sensors based on historical production data to form a correlation matrix; when the absolute value of the Pearson correlation coefficient between any two state parameters in the correlation matrix is ​​greater than a preset correlation threshold, it is determined that there is a strong correlation between the two state parameters. Specifically, to achieve a quantitative analysis of the inherent correlation between state parameters in the production process and to provide a theoretical basis for subsequent optimization decisions, a correlation analysis of state parameters needs to be performed. This analysis first requires obtaining sufficient historical production data covering different production stages and typical working conditions as the basis for analysis. Based on this data, the Pearson correlation coefficient between each pair of state parameters collected by sensors is calculated. This coefficient is obtained by comparing the covariance of the two parameters with the correlation coefficient between the two parameters. The ratio of the product of standard deviations is calculated, with a numerical range of [-1, 1]. The closer the absolute value is to 1, the stronger the linear correlation between the parameters. The correlation coefficients between all parameters are arranged in matrix form to construct a correlation matrix, thereby systematically representing the correlation pattern between parameters. The correlation matrix is ​​analyzed by a preset correlation threshold: when the absolute value of the Pearson correlation coefficient between any two state parameters is greater than the threshold, it is determined that there is a statistically significant strong correlation between them. This analysis result will serve as important prior knowledge to guide the parameter adjustment priority setting of subsequent optimization algorithms. Specifically, when real-time monitoring data indicates that a certain state parameter has significant fluctuations, the optimization algorithm can, based on the established strong correlation, prioritize targeted searches and fine adjustments in the parameter space that are strongly correlated with that parameter, thereby effectively improving the targeting and convergence efficiency of the optimization process.

[0061] The present invention is further configured such that the intelligent optimization model is a collaborative model of particle swarm optimization algorithm and long short-term memory network model; the initialization of the particle swarm optimization algorithm includes: setting a population containing a preset number of particles, each particle's position vector representing a combination of process parameters, and presetting a maximum number of iterations for the algorithm; the combination of process parameters includes the aggregate mix ratio of each cold aggregate bin, mixing time, drying drum burner temperature, and asphalt injection rate; the initialization of the long short-term memory network model includes: configuring its input layer to receive time-series data organized by a sliding time window of a specific time length, and configuring its output layer to... The method predicts key quality indicators (KPIs) for future batches of asphalt mixtures. The time-series data includes historical process parameters and real-time operating data. The KPIs include asphalt content, key sieve pass rate, porosity, and stability. The collaborative approach is achieved iteratively. Each iteration includes: generating a set of candidate process parameters using a particle swarm optimization algorithm based on the current population state; inputting each candidate process parameter set and real-time operating data into a long short-term memory (LSTM) network model to predict the KPIs of the asphalt mixture produced when applying the candidate process parameter sets; and using the deviation between the predicted KPIs and the target values ​​as the adaptation factor in the particle swarm optimization algorithm. The initialization process of the particle swarm optimization algorithm includes: firstly, setting a population of a preset number of particles, where the position coordinates of each particle in the search space correspond to a complete combination of process parameters. This combination specifically includes adjustable variables such as the aggregate mix ratio of each cold aggregate bin, mixing time, dry drum burner temperature, and asphalt injection amount. Simultaneously, a maximum number of iterations is set for the algorithm. The computational cost of the optimization process is controlled; the initialization process of the Long Short-Term Memory (LSTM) network model includes: configuring its input layer to receive time-series data organized by a sliding time window of a specific time length, such as corresponding to the past ten production cycles, which includes historical process parameter settings and corresponding real-time operating data; the network output layer is configured to predict multiple key quality indicators of future batches of asphalt mixture, including asphalt content, key sieve pass rate, porosity, and stability; the network can learn and memorize the dependencies across time steps in the production process through its internal gating mechanism;The collaborative optimization of the particle swarm optimization algorithm and the long short-term memory network model is achieved through an iterative process. The process is as follows: At the beginning of each iteration, the particle swarm optimization algorithm generates a set of candidate process parameters based on the position and velocity information of all particles in the current population. Then, each candidate set is combined with real-time operating data to form a complete input feature vector, which is input to the trained long short-term memory network model. This network predicts the key quality indicators of asphalt mixtures when using the candidate parameters based on the learned time dynamic characteristics. The particle swarm optimization algorithm converts the deviation between the predicted and target indicators into fitness values, and updates the velocity, position, and historical best position of each particle accordingly. After iterating until a preset termination condition is met, the process parameter combination corresponding to the historical best position of the population is output as the optimization result.

[0062] The present invention is further configured such that the digital twin model is a three-dimensional model of an asphalt mixing plant constructed based on building information modeling technology, and the digital twin model accurately maps the overall layout of the plant area and the structure of key equipment;

[0063] The digital twin model provides a detailed model of the mixing plant, drying drum, and dust removal system. It can display the flow state of materials inside the mixing tank and weighing bin, and dynamically simulate the rotational motion of the drying drum and the trajectory of the aggregate.

[0064] Each component in the digital twin model is associated with equipment attribute information and deeply integrated and dynamically mapped with real-time production data to achieve three-dimensional visualization and monitoring of production status. Specifically, the digital twin model is built based on Building Information Modeling (BIM) technology. Its implementation process begins with precise mapping and data collection of the mixing plant area. Based on this, a virtual digital model that completely corresponds to the physical entity is created using professional 3D modeling software. This model accurately maps the overall layout of the plant area, including the silo array, conveyor belt system, and main building structure, ensuring that the virtual spatial relationships are strictly consistent with the physical world. In terms of refined modeling of key equipment structures, not only is the external outline of the mixing tower constructed, but also high-precision geometric modeling is carried out on the internal core components such as the mixing cylinder, weighing silo, and hot material silo. The internal material flow status is visualized and displayed through the setting of transparent materials and dynamic particle effects. For the drying drum, the three-dimensional structure of the drum body and internal lifting plates is accurately constructed according to its mechanical design drawings. Animation technology or a physics engine is used to drive the dynamic simulation of the drum body's rotation and the trajectory of the aggregate under the action of the lifting plates. At the same time, the dust removal system's bag filter and flue are also constructed. A physical model containing the internal structure is established to simulate the flow and filtration process of dust airflow. Each component in the digital twin model is defined as an intelligent object, and its associated equipment attribute information, including equipment model, rated power, design life, and maintenance records, is stored in the asset performance management database and bound to the three-dimensional geometric components through a unique identifier, achieving deep integration of geometric and attribute information. The dynamic mapping between the model and real-time production data is achieved through data interfaces and driving logic. The system obtains data streams from the production control system and optimization algorithms in real time through industrial data interfaces. After parsing, the system drives the state of the corresponding components of the twin according to predefined rules. For example, the real-time process parameters of the mixing plant are displayed in the form of data tags at the corresponding positions in the model, and the operating parameters of the drying drum control the playback state of its three-dimensional animation. Based on real-time energy consumption data, a heat map is generated on the plant floor plan to display the energy consumption distribution. Based on the real-time calculated emissions, a particle flow emitted from the emission source is generated and controlled to simulate the virtual flow of carbon emissions. Thus, through the combination of precise geometric modeling, attribute information binding, and real-time data driving, three-dimensional visualization monitoring and mapping of all elements and states of the mixing plant are achieved.

[0065] The present invention is further configured such that the visualization monitoring includes process parameter transparency, production status visualization, and carbon emission visualization:

[0066] The process parameters are made transparent by displaying aggregate ratio, asphalt dosage, mixing time, inlet and outlet temperatures and rotation speed in real time on the corresponding equipment in the digital twin model;

[0067] The production status visualization is displayed on the plant layout as a heat map showing the distribution of global energy consumption or production efficiency.

[0068] The carbon emission visualization is based on real-time collected production data, dynamically calculating instantaneous carbon emissions, and simulating the carbon emission trajectory in a digital twin model using particle flow, visually displaying the emission source to the environment. Specifically, the visualization monitoring function in this embodiment is achieved through data integration, graphics mapping, and real-time rendering technologies: the transparency of process parameters relies on the deep integration of the digital twin model and real-time data streams. The data acquisition interface deployed in the mixing plant control system periodically acquires the current values ​​of key process parameters such as aggregate ratio, asphalt dosage, mixing time, inlet and outlet temperatures, and rotational speed. After these data are transmitted to the digital twin platform via a data bus, the platform kernel associates each parameter with the corresponding equipment component in the 3D model according to predefined mapping rules. The rendering engine dynamically generates information label layers when drawing the scene, displaying the parameter values ​​in real-time as text or digital instruments overlaid on the visible area of ​​the corresponding equipment model. The core of production status visualization lies in... Abstract production indicators are transformed into spatial distribution images. Energy consumption or efficiency data of different areas or equipment units throughout the plant are obtained from the energy management system and production execution system. After normalization by the data processing module, a corresponding color value is assigned to each data point according to the preset numerical range and color gradient mapping relationship. The graphics engine then uses the two-dimensional plan of the plant as the base map and fills the corresponding area with the calculated color according to the geographical location information of each unit to generate a heat map that can intuitively reflect the distribution of global indicators. Carbon emission visualization includes two key links: calculation and rendering. The platform's built-in carbon calculation model is based on real-time collected fuel consumption and electricity consumption data. Combined with pre-entered carbon emission factors of various energy sources, the instantaneous carbon emission is calculated through weighted calculation. The three-dimensional engine dynamically generates virtual particles starting from the emission source. The life cycle, trajectory and concentration of each particle are dynamically driven by the instantaneous carbon emission, thereby constructing a visualized carbon emission trajectory from the emission source to the environment in the three-dimensional scene.

[0069] The present invention is further configured such that the closed-loop control is implemented through a rule-based three-level alarm mechanism:

[0070] Level 1 warning: Triggered when process parameters deviate from the preset optimal value but are still within the preset safety range, the relevant equipment or area is highlighted in blue in the digital twin model;

[0071] Level 2 Alarm: Triggered when process parameters are greater than or equal to a preset safety threshold, automatically generating an early warning prompt and initiating the maintenance work order generation process;

[0072] Level 3 Alarm: Triggered when process parameters are greater than or equal to preset critical danger values. A red high-brightness flashing alarm is activated for the relevant equipment or area in the digital twin model, and automatic interlocking control is implemented, executing operations to reduce equipment load or initiate an emergency shutdown procedure. Specifically, the rule-based level 3 alarm mechanism achieves closed-loop control through the following process: Its data comparison function continuously compares the real-time collected process parameters with the optimal values ​​and safe ranges in the preset rule library. When a parameter deviates from the optimal value but remains within the preset safe range, a level 1 warning is triggered. The digital twin platform then sends a signal to the 3D rendering unit through its command sending function, causing a change in the visual representation parameters of the corresponding equipment material in the digital twin model. The system displays a blue highlight effect. When the process parameter is greater than or equal to the preset safety threshold, a level two alarm is triggered, automatically activating its warning interface generation function to pop up a warning window. Simultaneously, its work order processing logic is initiated. This logic calls the preset maintenance work order template based on the equipment identifier and abnormal parameter information, fills in the key content, forms a preliminary maintenance work order, and sends it to the maintenance scheduling system. When the process parameter is greater than or equal to the preset danger threshold, a level three alarm is triggered. The 3D rendering unit controls the relevant equipment model to display a red high-frequency flashing alarm state. At the same time, the safety interlock mechanism is activated, sending predefined safety instructions to the underlying programmable controller through its control signal output function to perform operations such as reducing the equipment operating load or initiating an emergency shutdown procedure.

[0073] The present invention is further configured such that the closed-loop control is specifically implemented in an adaptive adjustment process for abnormal asphalt viscosity conditions, including: when the online viscometer detects that the asphalt viscosity value exceeds the preset optimal range for a consecutive preset number of sampling periods, the particle swarm optimization algorithm is automatically triggered to recalculate the optimized process parameter settings with the current asphalt viscosity as the optimization boundary condition and maintaining the target porosity as the optimization objective; based on the recalculated optimized process parameter settings, an adjustment command is issued to the mixing plant control system; the present invention is further configured such that the adjustment command includes a compensation value for the temperature of the dry drum burner and an extension value for the wet mixing time of the mixing cylinder; specifically, when the online viscometer detects that the asphalt viscosity value continuously deviates from the preset optimal range within a consecutive preset number of sampling periods, the recalculation process of the particle swarm optimization algorithm is automatically triggered; this process uses the currently detected abnormal asphalt viscosity value as the new process constraint boundary condition and maintains the target porosity of the asphalt mixture as the core optimization objective, and reactivates the particle swarm optimization algorithm for iterative calculation. In this optimization calculation process, the algorithm's search space is strictly constrained by the current asphalt viscosity value. The main adjustment directions for improving asphalt fluidity are increasing aggregate heating temperature and extending mixing time. Candidate process parameter schemes generated by the algorithm are sequentially input into a pre-trained Long Short-Term Memory (LSTM) network prediction model for rapid evaluation. This model is based on real-time production data and the predicted mixture void ratio and other key quality indicators. The Particle Swarm Optimization (PSO) algorithm uses the deviation between the predicted void ratio and the target value as the fitness evaluation criterion. Through iterative optimization, it outputs a set of process parameter settings that optimally maintain the target void ratio under the new asphalt viscosity constraint. Based on the optimized process parameter settings, specific adjustment instructions are issued to the mixing plant control system. These instructions include the compensation value for the dry drum burner setting temperature calculated based on the viscosity deviation, used to improve the workability of high-viscosity asphalt by increasing aggregate temperature, and simultaneously extending the wet mixing time of the mixing drum to ensure a more complete mixing cycle for asphalt materials, thereby achieving uniform dispersion.

[0074] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twin, characterized in that, include: Multi-source sensor data are collected during the production process of asphalt mixing plant, and the multi-source sensor data is fused to obtain a set of state parameter features. The multi-source sensor data includes temperature, humidity, flow rate, vibration signal, dust concentration and asphalt viscosity. The state parameter feature set is input into the intelligent optimization model, and the optimized process parameter set value is output through iterative search and prediction feedback. The optimized process parameter settings are sent to the asphalt mixing plant control system for execution, and the digital twin model is driven by real-time production data for visual monitoring and closed-loop control.

2. The dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twin as described in claim 1, characterized in that, The fusion processing of multi-source sensor data includes: An extended Kalman filter is used to filter the acquired continuous variable sensor data to eliminate high-frequency noise; Principal component analysis was performed on the raw sensor data to extract the top K principal components whose cumulative variance contribution rate was greater than a preset contribution rate threshold. When the difference in readings between data from the same type of sensor exceeds a preset difference threshold, a data fusion algorithm is used to process the data and obtain consistent data. Based on historical production data, a correlation analysis of state parameters is conducted.

3. The dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twin as described in claim 2, characterized in that, The specific processing using the data fusion algorithm includes: A basic probability assignment function is constructed for each data source with reading differences. The assignment of the basic probability assignment function is based on the historical measurement accuracy of the data source and the degree of deviation of its current reading from the statistical characteristics of the reading group of other data sources in the same period. Based on the DS evidence theory, the basic probability assignment function is synthesized and calculated, and the consistency data is determined according to the maximum confidence proposition of the synthesis result.

4. The dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twin as described in claim 2, characterized in that, The correlation analysis of the state parameters includes: Based on historical production data, the Pearson correlation coefficients between each pair of state parameters collected by the sensors are calculated to form a correlation matrix; When the absolute value of the Pearson correlation coefficient between any two state parameters in the correlation matrix is ​​greater than the preset correlation threshold, it is determined that there is a strong correlation between the two state parameters.

5. The dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twin as described in claim 1, characterized in that, The intelligent optimization model is a collaborative model of particle swarm optimization algorithm and long short-term memory network model; The initialization of the particle swarm optimization algorithm includes: setting up a population containing a preset number of particles, where the position vector of each particle represents a combination of process parameters, and setting a maximum number of iterations for the algorithm. The combination of process parameters includes the aggregate mix ratio, mixing time, dry drum burner temperature, and asphalt injection amount for each cold aggregate bin. The initialization of the Long Short-Term Memory Network model includes: configuring its input layer to receive time-series data organized by sliding time windows of a specific time length, and configuring its output layer to predict key quality indicators of future batches of asphalt mixtures. The time-series data includes historical process parameters and real-time operating condition data, and the key quality indicators include asphalt content, key sieve pass rate, porosity, and stability. The collaboration is achieved iteratively, with each iteration including: The particle swarm optimization algorithm generates a set of candidate process parameters based on the current population state; Each candidate process parameter set and real-time operating data are input into a long short-term memory network model to predict the key quality indicators of the asphalt mixture produced when the candidate process parameter set is applied. The particle swarm optimization algorithm updates particle velocity, position, and the swarm's historical best position based on the deviation between the predicted key quality indicators and the target value. The iteration process continues until the preset termination condition is met, and the combination of process parameters corresponding to the group's historical best position is output as the optimized process parameter set value.

6. The method for dynamic optimization of asphalt mixing process based on multi-data fusion and digital twin as described in claim 1, characterized in that, The digital twin model is a three-dimensional model of an asphalt mixing plant built based on building information modeling technology. This digital twin model accurately maps the overall layout of the plant and the structure of key equipment. The digital twin model provides a detailed model of the mixing plant, drying drum, and dust removal system. It can display the flow state of materials inside the mixing tank and weighing bin, and dynamically simulate the rotational motion of the drying drum and the trajectory of the aggregate. Each component in the digital twin model is associated with equipment attribute information and is deeply integrated and dynamically mapped with real-time production data to achieve three-dimensional visualization monitoring of production status.

7. The dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twin as described in claim 1, characterized in that, The visualization monitoring includes transparency of process parameters, visualization of production status, and visualization of carbon emissions: The process parameters are made transparent by displaying aggregate ratio, asphalt dosage, mixing time, inlet and outlet temperatures and rotation speed in real time on the corresponding equipment in the digital twin model; The production status visualization is displayed on the plant layout as a heat map showing the distribution of global energy consumption or production efficiency. The carbon emission visualization is based on real-time collected production data, dynamically calculates instantaneous carbon emissions, and simulates the carbon emission trajectory in a digital twin model in the form of particle flow, visually displaying the emission source to the environment.

8. The dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twin as described in claim 1, characterized in that, The closed-loop control is achieved through a rule-based three-level alarm mechanism: Level 1 warning: Triggered when process parameters deviate from the preset optimal value but are still within the preset safety range, the relevant equipment or area is highlighted in blue in the digital twin model; Level 2 Alarm: Triggered when process parameters are greater than or equal to preset safety thresholds, automatically generating an early warning prompt and initiating the maintenance work order generation process; Level 3 alarm: Triggered when process parameters are greater than or equal to preset danger thresholds, the relevant equipment or area will be highlighted and flashed red in the digital twin model, and automatic interlock control will be implemented to reduce equipment load or start emergency shutdown procedures.

9. The method for dynamic optimization of asphalt mixing process based on multi-data fusion and digital twin as described in claim 1, characterized in that, The closed-loop control is specifically implemented in an adaptive adjustment process for abnormal asphalt viscosity conditions, including: When the online viscometer detects that the asphalt viscosity value exceeds the preset optimal range for a consecutive preset number of sampling periods, the particle swarm optimization algorithm is automatically triggered to use the current asphalt viscosity as the optimization boundary condition, maintain the target porosity as the optimization objective, and recalculate the optimized process parameter settings. Based on the recalculated optimized process parameter settings, adjustment instructions are sent to the mixing plant control system.

10. The dynamic optimization method for asphalt mixing process based on multi-data fusion and digital twin as described in claim 9, characterized in that, The adjustment instructions include a compensation value for the temperature of the dryer drum burner and an extension value for the wet mixing time in the mixing tank.

Images